White Papers

GAS COMPOSITION ANALYSIS OF HERMETIC STRUCTURES

AS A NON-DESTRUCTIVE TEST

Since 1967 the author has been engaged in the gas analysis of hermetically sealed structures. Historically this test procedure has been a destructive test and is specified as such in MIL-STD 883D, Method 1018.2, Procedure 1. During the past fifteen years experiments using a laser drill have been conducted to demonstrate the feasibility of analyzing gases in hermetic structures, backfilling the structure, and then resealing the structure. This paper presents a brief history of the evolution of the test procedure, comparison of the standard method verses the laser method and recommendation for implementation of the test method.

I. INTRODUCTION

The “RGA Test”, “Moisture Test”, or sometimes fondly referred to as the “Kiss of Death Test” is basically the procedure used to measure the amount of moisture as a percentage per unit volume contained in a hermetically sealed semiconductor package.

Test of this type, first conducted in 1967 by the author using a large double focused magnetic sector mass spectrometer, have been considered to be a destructive test enven when applied to DPA or failure analysis. Once the hermetic integrity of the package is breached in order to extract the gas sample, the package is usually discarded unless further physical or electrical examination is required.

As the value of the semiconductor device increases the motivation to reuse the package increases. The author had developed ESD handling procedures such that a high value package can be reprocessed after the moisture test but this procedure amounts to remanufacturing the device with very little savings in time or cost.

A considerable amount of interest and encouragement was received from RADC, TRW, Littion, JPL and Lockheed to proceed with experiments designed to reduce the cost of reprocessing.

The following experiments were designed to test the feasibility of analyzing the contents of a hermetic package using MIL-STD 883, Method 1018.2, Procedure 1, then backfilling the package with a clean dry gas and resealing the package. The basis for this work originated in 1975 when the author designed, manufactured and tested a similar system for Argonne National Laboratory to analyze fision gases in nuclear reactor fuel rods. This procedure consisted of:

  1. Fixturing the fuel rod and forming a vacuum tight seal to the surface,
  1. Laser drilling a hole to release the gas sample,
  1. Analyzing to gas sample,
  1. Backfilling the rod with a tag gas,
  1. Welding the hole closed with 95% material thickness in weld area
  1. Leak testing the weld area.

 

II. APPLICATION QUESTIONS

Applying the above procedure to semiconductor packages seems straight forward, however, a number of questions had to be answered, namely:

  1. Would laser drilling and/or rewelding created particulates which would impact the reliability and operation of the device under test?
  1. Can the various materials used to package semiconductor devices be laser drilled and rewelded reliably to insure hermeticity?
  1. Would laser drilling and/or rewelding damage the semiconductor devices?
  1. Would laser drilling impact the accuracy of the moisture measurement?
  1. What moisture level can be expected in the resealed package?
  1. Is the procedure economically viable?

III. EXPERIMENTS 

The following four experiments were conducted over the past two years with lots of encouragement and little or no funding from any of the parties involved: 1,test of particulate generation, reweldability, and circuit damage; 2, test reweldability, circuit damage, and performance of the resealed device; 3, test of moisture data accuracy; 4, test of moisture level in resealed package.

1. Experiment 1.

The author received two lids and three large hybrid packages (approximately 4.0cc in volume). Two of the sealed packages were empty and the third contained a scrap circuit on a large substrate which filled the entire package. Pin tests had been performed on all three packages as well as fine and gross leak tests. A series of 200 holes were drilled and rewelded on the two sample lids in order to optimize the laser operating parameters. Next six holes were drilled and rewelded in each of the three packages using the optimum settings on the laser. Pin test were then performed on the resealed packages. The packages were then opened and the rewelds and substrate were examined. The results of the pin tests and visual examination indicated that the procedure for drilling and rewelding would not create any particulates, the hermeticity could be maintained after the reweld, and the location of the penetration is critical to prevent damage to the circuitry.

2. Experiment 2.

The second experiment consisted of analyzing ten devices who's performance was particularly sensitive to high moisture levels. Five of the samples exhibited normal performance and five were electrically unstable. All ten devices were subjected to laser drilling, analysis, of the gas mixture, backfilling with grade 4.5 nitrogen, and rewelding. The results of this experiment indicated that those packages with high moisture levels were unstable and all rewelded packages exhibited normal performance.

3. Experiment 3.

The third experiment consisted of analyzing 96 packages (MC1 420GXKOREAJHBIK) randomly selected from a lot of 1200 packages, which were manufactured about 18 years ago for Pernicka Corporation. The packages were divided into 4 groups. The first group of 24 were to be analyzed using an external test fixture and mechanical puncture device. The second group of 24 were to be analyzed using a batch fixture and mechanical puncture device. The third and forth group of 24 each were to be laser drilled using an external fixture and batch fixture respectively. Each group was further divided in A and B subgroups of 12 packages which represented approximately one days work on the mass spectrometer.

Prior to each days analysis the mass spectrometer moisture calibration was checked using a high pressure saturator and General Eastern model 1500 which was calibrated by GEI against their NIST transfer standard. The uncertainty of the transfer standard was reported at +/-0.04 deg.C in measurement range of interest and our instrument difference reading varied from 0.02deg C to –0.08deg.C over the same range. Converting the above dewpoints to ppm per unit volume yields an accuracy of +/-18.4ppm –9.2ppm to –36.8ppm. A series of ten measurements were made each morning and the mean and standard deviation for each subgroup are summarized in the Table (1) below:

 

SUBGROUP DEWPOINT MEASURED MASS SPECTROMETER READING

1A 5280ppm+/-18ppm 5305ppm+/-25ppm

1B 5177ppm+/-25ppm 5160ppm+/-30ppm

2A 5220ppm+/-15ppm 5210ppm+/-27ppm

2B 5152ppm+/-20ppm 5151ppm+/-28ppm

3A 5210ppm+/-23ppm 5200ppm+/-22ppm

3B 5230ppm+/-20ppm 5243ppm+/-29ppm

4A 5212ppm+/-24ppm 5210ppm+/-30ppm

4B 5198ppm+/-26ppm 5205ppm+/-28ppm

Table 1. Subgroup Moisture Calibration Summary

After the ten calibration checks were made, 12 packages of each respective subgroups were analyzed per the requirements of MIL-STD 883D, Method 1018.2, Procedure 1. The results of these tests are summarized in Table (2) below:

SUBGROUP MEAN VALUE IN % STANDARD DEVIATION

1A 1.2229 .6398

1B 1.1531 .6523

2A 1.2374 .7511

2B 1.1003 .6887

3A 1.2577 .8012

3B 1.3049 .7212

4A 1.2168 .6656

4B 1.2446 .7143

Table 2. Subgroup Moisture Measurement Summary.

Careful review of Table (1) and Table (2) suggests that the packages selected for this experiment were by no means optimum. However, one might conclude that variations in the mean value measured in each subgroup were considerably less than the standard deviation and therefore acceptable under the MIL-STD.

4. Experiment 4.

The packages laser drilled in the above experiment were backfilled with Grade 4.5 Nitrogen after each package was analyzed except that those packages analyzed in the batch mode were also filled as a batch and then rewelded. The packages in subgroups 3A and 3B were remixed and randomly divided into 3A1 and 3B1. Similarly 4A1 and 4B1. Subgroups 3A1 and 4A1 were analyzed by mechanical puncture in the batch and external fixture mode respectively. Where as subgroups 3B1 and 4B1 were analyzed using the laser drill in the batch and external fixture mode respectively. The following Table (3) summarizes the test results of the rewelded packages:

SUBGROUP MEAN VALUE IN PPM STANDARD DEVIATION

3A1 2280 759

3B1 2850 612

4A1 2631 937

4B1 2174 453

Table 3. rewelded Subgroup Moisture Measurement Summary.

Examination of Table (3) might suggest that the packages under test benefited from the test procedure even though they would have been rejected as failures when first analyzed. The variation measured between subgroups is within the same order of magnitude as the calculated Standard Deviation and certainly less than the accuracy guidelines (+/-1000ppm) required by the MIL-STD.

 

IV. CONCLUSION

The laser drill and reweld procedure as applied to the moisture analysis test can be applied without apparent distortion of the test result, or damage to the performance of the device under test provided that extreme caution is exercised in selecting the location of the puncture site. The economic benefit of applying this technique to analyze packages and/or rework packages depends entirely on the cost of the original package and retesting requirements. Not all package configurations lend themselves to this procedure, however designers can incorporate the requirements of this procedure to new designs.

Back to Top